ChatGPT boss wants HQ in Europe

Sam Altman wants you to know he loves Europe.

The CEO of OpenAI, the maker of the artificial intelligence tool ChatGPT, spent last week touring the Continent, stopping in Spain, France, Poland, Germany and the United Kingdom. He was at once talking AI regulation with policymakers — he met national leaders Pedro Sánchez, Emmanuel Macron, Mateusz Morawiecki, Olaf Scholz, and Rishi Sunak — and scouting locations for an OpenAI European office.

“We really need an office in Europe,” Altman told POLITICO at a Paris event Friday. “We also just really want one.” Under the European Union’s upcoming Artificial Intelligence Act, companies with EU-based users would need a presence in the bloc, with national "supervisory authorities" in charge of implementing the regulation. The eventual choice of its HQ location will, therefore, determine which member country will oversee it when it comes to enforcing the AI Act.

Since its launch in November 2022, OpenAI’s ChatGPT — a bot able to create texts such as songs, scripts, articles and software based on written prompts — has caused both optimism and anxiety about what the rise of AI means for the future of humankind. While some have marveled at the tool’s prowess in creating computer code and streamlining office work, others fear that it could be used to generate troves of automated disinformation, manipulative content and biased material — or even put millions of people out of a job.

Nevertheless, politicians appear eager to host the world’s hottest AI lab. Opening the event, French Digital Minister Jean-Noël Barrot read a ChatGPT-generated description of Altman (“innovative, influential, visionary”) before pitching France as a “great AI country” — rattling off a list that included talent, abundant nuclear energy (to power the computers underpinning the AI), and cultural heritage among its assets.

In the U.K., where Altman also briefed national security personnel, a person familiar with his conversation with Sunak, who was granted anonymity to talk of high-level meetings, described the British prime minister as “deferential.”

Altman is still deliberating on where to house the new office. “If you had to pick just based on the most AI research talent, you’d pick France,” he told POLITICO. “But I’ve been super-impressed by the talent and energy everywhere.” OpenAI already has staff working in London, according to LinkedIn, and in September 2022 it created a U.K. subsidiary, according to the country’s business registry.

In Paris, Altman strove to quash reports, from Reuters, that OpenAI might leave the EU if the AI Act proved too onerous. “We plan to comply. We want to offer services in Europe,” Altman told the Parisian audience. “We just want to make sure we’re technically able to. And the conversations have been super-productive this week,” he added.

First floated by the European Commission in 2021, the AI Act would ban some uses of AI uses (such as social scoring and some instances of facial recognition) and impose stricter rules related to safety and oversight when it comes to sensitive AI applications considered "high-risk." On top of that, according to a version of the AI Act adopted earlier this month by lawmakers in the European Parliament, “generative” models such as ChatGPT — which can create new content, like text or photos — would have to disclose a summary of copyrighted materials used as training data.

The rule — which still needs to be agreed upon by representatives of the Commission and EU member countries — addresses worries from artists and publishers that AI firms might use their intellectual property without their consent or knowledge.

“That sounds like a great thing to ask for,” Altman told POLITICO. “But — due to the way these datasets are collected and the fact people have been copying data in different ways on different websites — to say I have to legally warrant every piece of copyrighted content in there is not as easy as it sounds.”

Altman thinks an easier way for creators to know if their work is being used would be based on whether their names appear in the prompts users give to an AI. “Every time you ask, ‘I want a song in the style of the Beatles,’ that would be clear,” Altman said.

All in all, however, Altman struck an optimistic tone about the AI Act and said he’d be happy to meet EU policymakers — despite skipping a planned Brussels stop in his tour. He told POLITICO that OpenAI would join the EU’s first "sandbox,” based in Spain, where AI companies will be able to test their regulatory compliance.

“It’s going to get to a good place,” he said. “Regulatory clarity will be a good thing.”

This article has been updated.